Guides · Updated August 2026
Does AI Culling Software Write Its Picks Back to Lightroom Classic?
Short answer: some tools do, some don't — and the ones that don't are not always upfront about it. The thing to look for has a name: the XMP sidecar. If a culling tool writes its picks as standard XMP files next to your RAWs, your ratings travel with your images and your catalog reads them like your own. If it keeps selections inside its own app, every hour of culling you do there is an hour that stays there.
What an XMP sidecar actually is
An XMP sidecar is a small text file, based on Adobe's open XMP standard, saved next to a RAW file — same name, different extension. It carries the frame's metadata: star rating, pick or reject flag, color label, keywords. Because that data sits beside the image in the open, any catalog software that reads the standard sees the same ratings — Lightroom Classic being the one most working photographers already live in. The RAW itself is never touched; sidecars are non-destructive by design.
That's the whole reason the question matters more than any feature list. Your catalog, your presets, your folder discipline are years of accumulated work — the workflow you shouldn't have to rebuild for a culling tool. A tool that speaks XMP joins that system. A tool that doesn't asks you to move into its house.
Where tools differ, quietly
Three patterns exist on the market. Some tools write standard sidecars as they rate — the cleanest case. Some export a selection only at the end, sometimes in a proprietary format that needs an import step. Some keep everything in their own database or cloud, and what you get out is a folder of files with the metadata stripped of the decisions you made. Product pages tend to say "integrates with your workflow" for all three — which is why the verification below is yours to run, and why one word in the marketing deserves attention: integrates can mean anything; writes XMP sidecars means one thing.
A second, subtler difference: direction. Writing picks out is half the job; reading the ratings you've already made is the other half — it's what lets a second pass build on a first one instead of flattening it. Terminology isn't standardized between apps (picks, flags, and color labels don't always map one-to-one), so even between two XMP-speaking tools, check what arrives on the other side.
The four checks, before any money moves
1. Standard sidecars, visibly. After rating a few frames on a trial, look in the folder: is there an .xmp file next to each RAW? A sidecar you can see in the file browser settles the question; an export button or a proprietary list leaves it open.
2. Direction. Does the tool also read existing ratings, or only write? If you've already flagged rejects on import, a write-only tool will ignore that work.
3. RAWs untouched. Sidecars are non-destructive by design — confirm the tool holds to that and doesn't rewrite or re-save your originals.
4. One real shoot, end to end. The only test that settles it: run a finished shoot through the trial, open your own catalog, and confirm the stars and flags are sitting where you expect. It's the catalog-side twin of auditing the picks themselves against your own eye — and it's one of the eleven checks in The Wedding Culling Audit →
See how BestTake culls a wedding the same day →
Further reading
This guide describes the open Adobe XMP standard and general market patterns; it makes no claim about any specific tool's implementation — including ours: run check 4 on anything you evaluate. Content written with AI assistance, reviewed before publication by the publisher.